Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The process followed in this study is clearly articulated, providing a comprehensive understanding of the steps involved. However, some aspects may benefit from simplification to enhance clarity for a broader audience. The analysis itself is robust, and the images and methodology are well-presented, making the procedure both followable and repeatable.

Despite the methodological strengths, the results leave much to be desired. A significant number of non-river pixels were incorrectly classified as river pixels, leading to substantial discrepancies between the real areas and the extracted values, as evidenced by the data in Tables 4 and 5.

From a hydrological perspective, the information provided is not usable due to these inaccuracies. While the study offers interesting insights into remote sensing, it falls short of the primary goal of improving image quality similar to Sentinel's standards.

It is recommended to explore the use of different water indexes or alternative images to potentially enhance the accuracy of the classifications. Adjusting these variables may help achieve more reliable results and better fulfill the study's objectives.

Overall, while the methodological approach is solid, the results undermine the study's utility and impact in its intended field.

 

Author Response

Reviewer 1:

The process followed in this study is clearly articulated, providing a comprehensive understanding of the steps involved. However, some aspects may benefit from simplification to enhance clarity for a broader audience. The analysis itself is robust, and the images and methodology are well-presented, making the procedure both followable and repeatable.

Response: Thank you for your recognition of our work. We have revised our manuscript according to your comments as follows:

 

Despite the methodological strengths, the results leave much to be desired. A significant number of non-river pixels were incorrectly classified as river pixels, leading to substantial discrepancies between the real areas and the extracted values, as evidenced by the data in Tables 4 and 5.

Response: Thank you so much for your detailed comments.

We have investigated the causes of those poor results in Table 4 and 5 and further discussed them in Section 5.2.

There are mainly two reasons making non-river pixels incorrectly classified as river:

The first one is the thresholding process. The threshold used in this study is determined based on confidence, which may be inaccurate at some river sections. Land pixels with reflectance value similar to water may also be included in the confidence interval, resulting in low classification accuracy.

The second one is the hill shade in the study area. The hill shade also would interfere river water mapping, leading to non-water pixels resemble water (Jia, N. 2008).

Meanwhile, we also proposed some possible strategy to deal with these issues in the future:

For the inaccurate threshold, a more robust determination method could be applied, such as using machine learning to improve confidence threshold (Smets, L. et al. 2023). For the hill shade, terrain information such as HAND index (Van Der Meer, F.D. t al. 2014) could be involved into river extraction framework, to further distinguish the hill shade and river pixels which have similar reflectance. Moreover, image fusion is also an effective method to reduce the uncertainties. (Almarines, N.R. et al. 2024).

 

From a hydrological perspective, the information provided is not usable due to these inaccuracies. While the study offers interesting insights into remote sensing, it falls short of the primary goal of improving image quality similar to Sentinel's standards.

It is recommended to explore the use of different water indexes or alternative images to potentially enhance the accuracy of the classifications. Adjusting these variables may help achieve more reliable results and better fulfill the study's objectives.

Overall, while the methodological approach is solid, the results undermine the study's utility and impact in its intended field.

Response: Thank you for pointing this out. The aim of this article is to evaluate the ability of the SDGSAT-1 satellite in river water mapping at Qinghai Tibet Plateau. This has now been further emphasized in the Introduction:

To evaluate the performance of SDGSAT-1 data in extracting water bodies, a comparative analysis was conducted on the Dang River located on the Qinghai-Tibet Plateau (QTP), employing NDWI, MDNWI, and SWI indices for accurate extraction.

There are mainly two reasons when selecting these three indices:

Firstly, we considered the band setting and resolution characteristics of the SDGSAT-1 satellite, which matched well with these three water indices.

Secondly, NDWI works well for open areas like lakes and large rivers. MNDWI, does well in eliminating shadows from mountainous rivers and water bodies. SWI helps distinguish water from shadows, especially in complex terrain like mountains. For rivers like Dang River with a wide watershed and located in mountainous areas, these three indices are representative ones that deserve investigation.

However, due to the limitation of the number of bands, there are not as many water indices as Sentinel-2 that can be used, which is also a future research direction to explore some improved water indices (Chen, J. et al. 2024). We also proposed some possible research methods: like using machine learning to optimize water indices (Smets, L. et al. 2023), taking into account the characteristics of the water body and optimizing the water indices by considering factors like turbidity, or proposing water indices that are more suitable for the SDGSAT-1 satellite based on the characteristics of its satellite sensors.

This has now been clarified in Section 5.2.

 

 

References：

Jia, N. Research on the Protection and Sustainable Utilization of Water Resources in the Dang River Basin. 2008.

 

Almarines, N.R.; Hashimoto, S.; Pulhin, J.M.; Tiburan, C.L.; Magpantay, A.T.; Saito, O. Influence of Image Compositing and Multisource Data Fusion on Multitemporal Land Cover Mapping of Two Philippine Watersheds. Remote Sensing 2024, 16, 2167, doi:10.3390/rs16122167.

 

Chen, J.; Wang, Y.; Wang, J.; Zhang, Y.; Xu, Y.; Yang, O.; Zhang, R.; Wang, J.; Wang, Z.; Lu, F.; et al. The Performance of Landsat-8 and Landsat-9 Data for Water Body Extraction Based on Various Water Indices: A Comparative Analysis. Remote Sensing 2024, 16, 1984, doi:10.3390/rs16111984.

 

Smets, L.; Van Leekwijck, W.; Tsang, I.J.; Latré, S. Training a Hyperdimensional Computing Classifier Using a Threshold on Its Confidence. Neural Computation 2023, 1–18, doi:10.1162/neco_a_01618.

 

Van Der Meer, F.D.; Van Der Werff, H.M.A.; Van Ruitenbeek, F.J.A. Potential of ESA’s Sentinel-2 for Geological Applications. Remote Sensing of Environment 2014, 148, 124–133, doi:10.1016/j.rse.2014.03.022.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

1. Which images are the background for the study area? The background images can’t show the characteristics of the river clearly.   

2. In the subtitle 3.2, ROI was used in the title, however, AOI was used in the text. It is recommended to unify them. 

3. From the results, the accuracy of water bodies extraction is not high. What is the reason for the misclassification?  Whether different thresholds can be set for different sections of rivers?

4. There are some spellings and grammatic errors.  Such as Line 13 in the abstract section “research This study evaluates the effectiveness of SDGSAT-1 in extracting water…”; Line 348, “finesse This advantage mainly stems from…”. 

Author Response

Reviewer 2:

1. Which images are the background for the study area? The background images can’t show the characteristics of the river clearly.

Response: Thank you so much for your helpful and detailed comments. We have made the following revisions to Figure 1:

1. Using brighter colors to represent rivers.
2. Enhancing the contrast of the background image.

Figure 1. Schematic illustration of study area: (a) location of the Dang River; (b-d) three test sites corresponding to A, B, and C in (a).

 

2. In the subtitle 3.2, ROI was used in the title, however, AOI was used in the text. It is recommended to unify them.

Response: Thanks for the very helpful comments. We have replaced AOI with ROI in the whole manuscript.

 

3. From the results, the accuracy of water bodies extraction is not high. What is the reason for the misclassification? Whether different thresholds can be set for different sections of rivers?

Response: Thank you for pointing this out. We have conducted a further analysis of the misclassification and added a detailed explanation in Section 5.2. River extent extraction at our study area could be easily affected by the hill shade (Jia, N. 2008) and threshold setting. Meanwhile, we also proposed some possible strategy to deal with these issues in the future. For the inaccurate threshold, a more robust determination method could be applied, such as using machine learning to improve confidence threshold (Smets, L. et al. 2023).

In this study, we mainly considered the ability of the SDGSAT-1 satellite when setting the confidence threshold for the whole river. Therefore, setting thresholds for different river sections will be conducted in our future work.

 

4. There are some spellings and grammatic errors. Such as Line 13 in the abstract section “research This study evaluates the effectiveness of SDGSAT-1 in extracting water…”; Line 348, “finesse This advantage mainly stems from…”.

Response: Spellings and grammatic errors have been carefully examined and corrected in the whole manuscript.

 

 

References：

Jia, N. Research on the Protection and Sustainable Utilization of Water Resources in the Dang River Basin. 2008.

 

Smets, L.; Van Leekwijck, W.; Tsang, I.J.; Latré, S. Training a Hyperdimensional Computing Classifier Using a Threshold on Its Confidence. Neural Computation 2023, 1–18, doi:10.1162/neco_a_01618.

Author Response File: Author Response.pdf